18th International Conference on Electricity Distribution
CIRED
Turin, 6-9 June 2005
STRATEGIES FOR DFIG VOLTAGE CONTROL DURING TRANSIENT OPERATION Mustafa KAYIKCI, Olimpo ANAYA-LARA, Jovica V. MILANOVIC, Nick JENKINS Manchester Centre for Electrical Energy, The University of Manchester – United Kingdom
[email protected] [email protected] [email protected] [email protected]
SUMMARY This paper explores and compares the performance of alternative voltage control strategies applied to DFIG. Different combinations of reactive power control of rotorand grid-side converters are investigated for voltage control purposes. Simulations are performed using detailed models built in DIgSILENT PowerFactory in order to illustrate the influence of controllers on transient stability of the wind turbine. Operational limits, such as current margins and PWM modulation limits are also taken into account.
power from the wind. Grid-side converter (GSC) can also be utilised for voltage control since it is capable of generating reactive power. (The generation of Q is limited by GSC rating.)
INTRODUCTION The fast development of wind power generation brings new requirements for wind turbine integration into the network. One of the new challenging issues is that a wind farm has to provide fault ride through capability and remain connected during network faults. Over the last few years doubly fed induction generator (DFIG) wind turbines have been the preferred option for high-capacity wind farms. This is due to their ability to control active power flow and reactive power exchange with the network, thus providing superior performance concerning system stability during large disturbances. Currently many operators prefer unity power factor operation since it is the active power production that is rewarded. Reactive power is produced only if there are sufficient financial incentives. However, as the penetration of the wind farms into the power system increases, ancillary services, such as voltage control, provided by the wind turbines become more significant. This paper explores various methods of voltage control that can be provided by a DFIG. MODELLING OF THE DFIG SYSTEM DFIG wind turbines are based on wound-rotor induction machines where the rotor circuit is fed through back-to-back voltage source converters. A typical configuration is shown in Fig.1. The model is built in DIgSILENT PowerFactory and further details can be found in [1, 2]. Active and reactive power output of the machine can be regulated independently, thanks to the four-quadrant operation of converters, by using suitable control reference frames (e.g. stator-flux orientation). This is a significant advantage since the reactive power (Q) can be regulated independently for voltage control purposes. Active power (P) is controlled such that the turbine extracts the maximum CIRED2005 Session No 4
Fig.1 Typical DFIG network connection
REACTIVE POWER CONTROL Neglecting the stator resistance and with a stator-flux vector orientation, the relationship between the total stator reactive power and the d-axis rotor current i dr is given by [3]: Qs =
Lm Vs
( Ls + Lm )
⋅ idr −
Vs
2
ω s ( Ls + Lm )
(1)
It can be seen that the production of Q is directly proportional to i dr , so by controlling i dr the stator reactive power production and consequently the voltage at DFIG terminals can be regulated. Fixed power factor operation This mode is usually employed as unity power factor (pf) operation (zero reactive power output). Other power factor values can be specified (e.g. from 0.95 leading to 0.95 lagging) as long as the thermal limits are not breached. In unity power factor mode, the rotor excitation is arranged such that all magnetising current is supplied through the rotor and stator behaves as reactive neutral (Qs=0). The reactive current supplied to the rotor by the rotor-side converter (RSC) is independent of the network. The rotor circuit and the grid are
18th International Conference on Electricity Distribution
CIRED
and very small amount is left to generate Q (95 kVAr). In order to generate the left half of the PQ curve, stator is set to absorb the maximum possible Q and the same procedure is applied as before. The dashed line in Fig.3 represents the prime mover limit corresponding to 1.27 MW stator active power. However, Fig.3 is only valid for nominal voltage and frequency. If voltage falls below nominal value, electrical torque decreases and more q-axis current is required to keep it at the reference value. This further reduces the margin for daxis rotor current to generate Q. The operating region indicated in Fig.3 would shrink from both (left and right) sides if the voltage was smaller than nominal.
DIgSILENT
decoupled by the dc-link in between. RSC, being a voltage source converter, can separately generate the reactive power needs of the rotor, provided that the currents are within converter limits and harmonics are acceptable (no over- or under-modulation). Reaching the voltage or current limits of the rotor during unity pf operation is unlikely since the maximum rotor voltage for the DFIG used in this study is ~410V (modulation index, Pm=0.56), which is below the modulation index limits, Pm=±1. The same applies for current limitation because the magnetising current supplied through the rotor is ~0.1 p.u and is independent of rotor speed.
Turin, 6-9 June 2005
1.0000
0.600
VOLTAGE CONTROL STRATEGIES
0.200
(a) -0.2000
Voltage control by rotor excitation
-0.6000
13.200
22.400
31.600
40.800
[s]
50.000
0.980
0.784
0.588
(b)
0.392
0.196
0.000 4.0000
13.200
22.400
31.600
40.800
[s]
50.000
22.400
31.600
40.800
[s]
50.000
Current Control: Stator-flux referred q-axis current (p.u.)
0.800
0.680
0.560
(c)
0.440
0.320
0.200 4.0000
13.200 DFIG: Rotor Current, Magnitude of Phasor in kA
Fig.2 Rotor current response to a linearly increasing wind speed. (a) d-axis rotor current (p.u.) for maximum Q generation (solid line) and for maximum Q absorption (dashed line)
Fig.2 and Fig.3 show the operating region of DFIG. This is obtained by setting the turbine to generate minimum P (at cutin speed) and absorb/generate maximum Q such that the rotor current is kept at 1 p.u and wind speed is increased linearly up to the rated speed.
In this scheme reactive power is varied in order to maintain the voltage at the point of common coupling within given reference values (in this case voltage is controlled at 1p.u.). The difference between the measured and reference signal is fed through an anti-windup P-I controller and reactive power reference is obtained, which is then sent to the power controller of DFIG. In some previous studies (e.g., [4]) P-I controller limits were set to ±0.5 p.u. resulting in a power factor of ~0.9 at rated operation. However, during the low active power generation, reactive power generation capacity is higher than 0.5 p.u. (see Fig. 3) so by setting such controller limits Q generation capability of the machine is unnecessarily limited. Therefore reactive power limits in this study are set to ±1 p.u. 1.125
DIgSILENT
-1.0000 4.0000
1.50
1.000 1.00
0.875
(a)
0.50
(b)
0.750
0.00
DIgSILENT
0.625
1.50
-0.976 Mvar
0.095 Mvar
-0.50 0.500
-1.00
0.375
Prime mover limit Y = 1.270 MW
0.00
1.20
1.00
2.00
[s]
3.00
0.80
0.00
1.00
2.00
[s]
3.00
0.00
1.00
2.00
[s]
3.00
1.19
0.60 1.18
0.90
0.40 1.17
Ps
(d)
0.20
(MW)
(c)
0.60
1.16
0.00
1.15 -0.20
0.30
-0.40
1.14 0.00
1.00
2.00
[s]
3.00
Fig.4 DFIG response to a voltage dip with voltage control by RSC. Solid – Active current prioritised, Dashed – Reactive current prioritised. All in p.u. (a) Voltage, (b) P stator, (c) Q stator, (d) Speed
0.00
-0.30 -1.00 x-Axis:
DFIG: Total Reactive Power/Terminal AC in Mvar DFIG: Maximum Q generation DFIG: Maximum Q absorption
Stator Q (MVAr)
0.00
1.00
Fig.3 Operating region of DFIG connected to an infinite bus
Active component of current (q-axis) is prioritised and as soon as the wind speed (or stator P) is increased, q-axis rotor current is increased leaving less space for d-axis current to produce reactive power and eventually Qstator starts to fall as seen in Fig.3. When the rated power is reached DFIG is using most of the rotor current on q-axis (i.e. iqr=0.99 p.u) CIRED2005 Session No 4
Fig.4 shows the voltage control action through rotor excitation. At 0.5s voltage control is activated and Q stator is increased to pull the voltage up from 0.98p.u. to 1p.u. At 1s, 3-phase fault is applied for 500ms and the voltage drops to 0.52p.u. Rotor current does not increase too much and crowbar protection is not activated. Since torque producing current (iqr) is prioritised, reactive current idr is set to zero (thus machine is partly excited from the stator) while iqr is boosted up to 1p.u. However, this is still not enough and the stator active power drops from 0.80p.u. to 0.56p.u. and rotor
18th International Conference on Electricity Distribution
CIRED
accelerates slightly. Alternatively the reactive current, idr, can be prioritised and voltage can be improved during the disturbance by ~6%. However this leads to uncontrolled active power reduction in order to reach the reference Q value. Q is boosted and P drops to zero and rotor starts to accelerate. Mechanical power can be limited by increasing the pitch angle in order to prevent over-speeding. However recent grid codes [5] only allow active power reduction proportional to the voltage drop. Moreover load distribution, fault impedance and type, X/R ratio of the network, and strength of the grid can increase the significance of real power generation for maintaining the voltage profile during the disturbance. Voltage control through grid-side converter
1.05
0.00
1.00
-0.25
0.95
-0.50
0.90
-0.75
0.85
-1.00
DIgSILENT
A GSC operates in an ac voltage-oriented reference frame. dand q-axis current components control the active and reactive power, respectively. d-axis current ensures that the rotor power is supplied/absorbed into/from the network so that dc link voltage is kept constant at 1 p.u. q-axis current (iq) can be varied in order to control terminal voltage by adjusting the reactive power output of the GSC. During normal operation a GSC is already partly-loaded by the transfer of rotor (slip) power. This may limit its reactive current generation. On the other hand, a GSC is rated for the maximum slip power transfer and most of the time turbine operates below rated power. Hence, rather than operating GSC at unity power factor idly, benefit should be taken from its available reactive current capacity.
Turin, 6-9 June 2005
Voltage control using both RSC and GSC There is very limited information in the literature about this voltage control scheme. Collaborative operation of RSC and GSC by trying to minimise the losses in both converters is discussed in [6]. Another method where GSC voltage control is activated only if the RSC is blocked due to heavy disturbances is proposed in [7]. In this paper two control strategies involving RSC and GSC are considered; the uncoordinated and coordinated operation of converters. a) Uncoordinated. When reactive power of RSC and GSC are both regulated for voltage control purpose, there is a risk that circulating currents may occur between the stator and GSC. Each of the two controllers is trying to regulate the voltage without ‘knowing’ what the other one is doing. This is illustrated in Fig.6 where the voltage control through GSC is activated at t=0.2s which pulls the voltage from 0.98p.u. to 1p.u. by increasing GSC reactive power. Later at 1s, the RSC voltage control is also activated to regulate the stator reactive power. However RSC control is faster compared to GSC control and it takes over and more importantly causes GSC to absorb significant amount of reactive power. In order to prevent the circulating currents, one of the voltage controllers can be operated with a droop characteristic. Bottom half of Fig.6 shows the results where GSC is operated with a droop characteristic with the same sequence of events. In this case at t=0.2s the voltage is not controlled to be 1p.u. instead, according to the droop, it is increased to 0.99p.u. When the RSC voltage control is activated at 1s, it takes over control in a smooth manner and GSC reactive power reduces to zero. Uncoordinated
GSC control operated with droop 1.02
Volatge (p.u.)
Volatge (p.u.)
1.02 1
0.98 0.96
6.4000
[s]
8.0000
-1.25 0.000
0.80
1.20
0.60
1.10
0.40
1.00
0.20
0.90
0.00
-0.20 0.000
1.6000
3.2000
4.8000
6.4000
[s]
3.2000
4.8000
6.4000
[s]
8.0000
0.70 0.000
1.6000
3.2000
4.8000
6.4000
[s]
8.0000
Fig.5 shows the voltage control action of GSC when the load is increased in steps of 20% every 0.5s. In response to the voltage drop, power controller acts fast and increases iq in order to increase QGSC. As the load is increased further, eventually the current limit of the GSC (i.e. ±1 p.u.) is reached and the voltage can not be controlled further and it starts to decrease. As the voltage drops below 0.95p.u., QGSC also starts to fall visibly while reactive current injection is kept at maximum. Reactive power output of GSC decreases linearly with the voltage (as opposed to the quadratic relationship for static VAR compensators.)
Session No 4
1
1.5
2
0 -0.2
0.5
1
1.5
2
0
0.5
1
1.5
2
0
0.5
1 time (s)
1.5
2
0.2 0
-0.2
0
0.5
1
1.5
2
1 0.5 0
-0.5
0 0.4
0.2
-0.4
Fig.5 GSC response to increasing load. (a) Voltage (p.u), (b) GSC q-axis current (p.u), (c) GSC reactive power (MVA), (d) PWM index
CIRED2005
0.5
0.4
0.80
1.6000
0.96
0
8.0000
GSC Q (MVAr)
4.8000
-0.4
Stator Q (MVAr)
3.2000
GSC Q (MVAr)
1.6000
Stator Q (MVAr)
0.80 0.000
1
0.98
0
0.5
1 time (s)
1.5
2
0.6 0.4 0.2 0
-0.2
Fig.6 Uncoordinated voltage control. Solid – Without droop, Dashed – With droop
b) Coordinated. Rather than using two different voltage controllers, the reactive power value can be split between the stator and GSC in a controlled manner. Different schemes can be adopted: (i) RSC can be made the ‘default’ controller and GSC can be used as a supplementary source only during certain circumstances. For example, GSC can be activated when the terminal voltage
